{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2021:XLMJAUZ7USXWLNYLATHQTVYWWD","short_pith_number":"pith:XLMJAUZ7","canonical_record":{"source":{"id":"2112.11668","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-22T05:04:41Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"fd4a50b3dcd869434d9d99a555f68c0a91bfccf1ec92771eea44dc8be47c4412","abstract_canon_sha256":"cd39e0db71b4ab3c8a5a04e73e95e08aa45cb1b6a5bf5a6210450d8ba6015a19"},"schema_version":"1.0"},"canonical_sha256":"bad890533fa4af65b70b04cf09d716b0d257790395667acc1b00a13413820f66","source":{"kind":"arxiv","id":"2112.11668","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2112.11668","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"arxiv_version","alias_value":"2112.11668v1","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.11668","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_12","alias_value":"XLMJAUZ7USXW","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_16","alias_value":"XLMJAUZ7USXWLNYL","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_8","alias_value":"XLMJAUZ7","created_at":"2026-07-05T03:43:09Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2021:XLMJAUZ7USXWLNYLATHQTVYWWD","target":"record","payload":{"canonical_record":{"source":{"id":"2112.11668","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-22T05:04:41Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"fd4a50b3dcd869434d9d99a555f68c0a91bfccf1ec92771eea44dc8be47c4412","abstract_canon_sha256":"cd39e0db71b4ab3c8a5a04e73e95e08aa45cb1b6a5bf5a6210450d8ba6015a19"},"schema_version":"1.0"},"canonical_sha256":"bad890533fa4af65b70b04cf09d716b0d257790395667acc1b00a13413820f66","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:43:09.483498Z","signature_b64":"Dh7gFY2Pi0N9PEKnJLfMDdKiaqC8KHYX+8AfoGLkMm2aDlzjteuC7k4N5H7T5sux/4GYdKGhgADzGyns4XAKBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bad890533fa4af65b70b04cf09d716b0d257790395667acc1b00a13413820f66","last_reissued_at":"2026-07-05T03:43:09.483149Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:43:09.483149Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2112.11668","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T03:43:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+zP1U9eo70k4jgzWAKzCkm1YUoiKx/HLSlXdet2Py/pZvKSoB/pWqw9XMezAv2dsChYYVdRZeuEOSqYdko/MCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T16:41:40.109587Z"},"content_sha256":"671f10d4af31a9c9e901f34b34d11aac156fe92fa93517bd790d287854ac8f9b","schema_version":"1.0","event_id":"sha256:671f10d4af31a9c9e901f34b34d11aac156fe92fa93517bd790d287854ac8f9b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2021:XLMJAUZ7USXWLNYLATHQTVYWWD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How Should Pre-Trained Language Models Be Fine-Tuned Towards Adversarial Robustness?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Hanwang Zhang, Luu Anh Tuan, Min Lin, Shuicheng Yan, Xinhsuai Dong","submitted_at":"2021-12-22T05:04:41Z","abstract_excerpt":"The fine-tuning of pre-trained language models has a great success in many NLP fields. Yet, it is strikingly vulnerable to adversarial examples, e.g., word substitution attacks using only synonyms can easily fool a BERT-based sentiment analysis model. In this paper, we demonstrate that adversarial training, the prevalent defense technique, does not directly fit a conventional fine-tuning scenario, because it suffers severely from catastrophic forgetting: failing to retain the generic and robust linguistic features that have already been captured by the pre-trained model. In this light, we prop"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.11668","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2112.11668/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T03:43:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1cRdtuo8vCyiSNCKUmKM7ij6iW82bqeYYc6pVhKhfrMGlf/kLW5mf8KqF3GAfKZ5v9LNPUYHePj2VZSmcLj9Cw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T16:41:40.111235Z"},"content_sha256":"598ff3c0b3aa3124866efc1022011bedfad2a38c766c5f4340a5627942a92802","schema_version":"1.0","event_id":"sha256:598ff3c0b3aa3124866efc1022011bedfad2a38c766c5f4340a5627942a92802"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/bundle.json","state_url":"https://pith.science/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-06T16:41:40Z","links":{"resolver":"https://pith.science/pith/XLMJAUZ7USXWLNYLATHQTVYWWD","bundle":"https://pith.science/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/bundle.json","state":"https://pith.science/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/XLMJAUZ7USXWLNYLATHQTVYWWD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:XLMJAUZ7USXWLNYLATHQTVYWWD","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"cd39e0db71b4ab3c8a5a04e73e95e08aa45cb1b6a5bf5a6210450d8ba6015a19","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-22T05:04:41Z","title_canon_sha256":"fd4a50b3dcd869434d9d99a555f68c0a91bfccf1ec92771eea44dc8be47c4412"},"schema_version":"1.0","source":{"id":"2112.11668","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2112.11668","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"arxiv_version","alias_value":"2112.11668v1","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.11668","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_12","alias_value":"XLMJAUZ7USXW","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_16","alias_value":"XLMJAUZ7USXWLNYL","created_at":"2026-07-05T03:43:09Z"},{"alias_kind":"pith_short_8","alias_value":"XLMJAUZ7","created_at":"2026-07-05T03:43:09Z"}],"graph_snapshots":[{"event_id":"sha256:598ff3c0b3aa3124866efc1022011bedfad2a38c766c5f4340a5627942a92802","target":"graph","created_at":"2026-07-05T03:43:09Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2112.11668/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The fine-tuning of pre-trained language models has a great success in many NLP fields. Yet, it is strikingly vulnerable to adversarial examples, e.g., word substitution attacks using only synonyms can easily fool a BERT-based sentiment analysis model. In this paper, we demonstrate that adversarial training, the prevalent defense technique, does not directly fit a conventional fine-tuning scenario, because it suffers severely from catastrophic forgetting: failing to retain the generic and robust linguistic features that have already been captured by the pre-trained model. In this light, we prop","authors_text":"Hanwang Zhang, Luu Anh Tuan, Min Lin, Shuicheng Yan, Xinhsuai Dong","cross_cats":["cs.LG"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-22T05:04:41Z","title":"How Should Pre-Trained Language Models Be Fine-Tuned Towards Adversarial Robustness?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.11668","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:671f10d4af31a9c9e901f34b34d11aac156fe92fa93517bd790d287854ac8f9b","target":"record","created_at":"2026-07-05T03:43:09Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"cd39e0db71b4ab3c8a5a04e73e95e08aa45cb1b6a5bf5a6210450d8ba6015a19","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-12-22T05:04:41Z","title_canon_sha256":"fd4a50b3dcd869434d9d99a555f68c0a91bfccf1ec92771eea44dc8be47c4412"},"schema_version":"1.0","source":{"id":"2112.11668","kind":"arxiv","version":1}},"canonical_sha256":"bad890533fa4af65b70b04cf09d716b0d257790395667acc1b00a13413820f66","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"bad890533fa4af65b70b04cf09d716b0d257790395667acc1b00a13413820f66","first_computed_at":"2026-07-05T03:43:09.483149Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T03:43:09.483149Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Dh7gFY2Pi0N9PEKnJLfMDdKiaqC8KHYX+8AfoGLkMm2aDlzjteuC7k4N5H7T5sux/4GYdKGhgADzGyns4XAKBg==","signature_status":"signed_v1","signed_at":"2026-07-05T03:43:09.483498Z","signed_message":"canonical_sha256_bytes"},"source_id":"2112.11668","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:671f10d4af31a9c9e901f34b34d11aac156fe92fa93517bd790d287854ac8f9b","sha256:598ff3c0b3aa3124866efc1022011bedfad2a38c766c5f4340a5627942a92802"],"state_sha256":"dfe3dc74dcb9d6c6ba4165ff1eb118aff2c1e68de2dc8f003f1a81d484719a02"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d3oIBOaLV7nnelnNHx3qExPl09DATqS1SxpDSX4qYD70xtca7C36e3lGOeFYLHmtGpi7MDSr404oj55zfMs0BA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T16:41:40.124354Z","bundle_sha256":"3d738e55a15085d5e9393294a81df07e0c89a090678e1495c6a71f32f208b1c5"}}